Remove outlier and fit curve

Hi all, I am quite new in the field of programming. I have a set of data with outliers, can someone please suggest a way of removing outliers and make the curve follow a normal trend of values instead of plotting NaN. Please see attached shot.

[KSSV]

No shot is attached...

[Star Strider]

We also need to see the code you are using to fit your data.

[Justin Igwe]

Sorry I have just attached the screen shot

Answer 1

[KSSV]

From the graph, I understand that, you want to remove the data which is greater then 23 (approximately). So, from the data get the position of values which are greater then 23 (your desired values) and make them NaN.

% Pseudo code 
  x = 1:100 ;
  y = rand(size(x)) ;
  idx = y>0.7 ;   % Get positions in y which are greater then 0.7 (~23 in your case)
  y(idx) = NaN ; 
  plot(x,y) ;

[Justin Igwe]

Hi Dr.Siva, thank you for your help. I actually needed to fill in the range with an estimated value as I do not want to have empty data such as NaN. So I used a code liike this:

clear all
clc
%max_fft_amplitudes_c1 = randn(315,1);
max_fft_amplitudes_c1=linspace(1,20,315);
for a=200:250
    max_fft_amplitudes_c1(a)=a/100*max_fft_amplitudes_c1(a);
end
max_fft_amplitudes_c1_clean = max_fft_amplitudes_c1;
figure  ('Name','mean_c1 vs mean_c1_clean','NumberTitle' ,'off')
plot (max_fft_amplitudes_c1)
hold on
%go through the whole data and find difference between current data and the
%future data
for b=1:314;
    c(b)=abs(max_fft_amplitudes_c1_clean(b) - max_fft_amplitudes_c1_clean(b+1));
   if c(b)>0.2 
       %if this difference is greater than a threshhold, reduce it by adding
       %0.06 to the previous data
        if b>3
            max_fft_amplitudes_c1_clean(b)=max_fft_amplitudes_c1_clean(b-1)+0.06;
        end
    end
 end
plot(c)
plot (max_fft_amplitudes_c1_clean)
    legend('max_fft_amplitudes_c1','c','max_fft_amplitudes_c1_clean')